Research Article

An Improved Unsupervised Single-Channel Speech Separation Algorithm for Processing Speech Sensor Signals

Table 2

The results of different speech separation methods (SNR, PESQ, STOI).

Voice
MethodIndex

MIXSNR1.645.45-0.771.83-0.610.08-2.07
PESQ2.272.052.431.581.841.751.73
STOI0.870.870.740.760.750.780.60

CNMFSNR9.317.938.468.525.946.384.89
PESQ2.852.212.491.851.991.972.08
STOI0.850.850.770.800.780.780.73

CNMF + ICASNR9.2810.625.426.887.195.712.63
PESQ2.101.841.801.781.801.661.78
STOI0.940.930.890.890.850.880.62

CNMF + JADESNR13.109.2011.198.447.038.325.68
PESQ3.022.402.692.052.202.262.35
STOI0.950.900.920.880.850.890.74